Yes. No law society bans it — what they regulate is competence, confidentiality, responsibility for the output and candour with the court.
Key takeaways
No Canadian law society prohibits lawyers from using artificial intelligence. What the regulators govern is how you use it, and the obligations they apply are not new ones written for AI: competence, confidentiality, professional responsibility for the work, and candour with the court.
So the honest answer to "can lawyers use AI in Canada" is that the question was settled some time ago and replaced by a harder one — what you must do before, during and after using it.
Ontario’s Rules of Professional Conduct require a lawyer to perform legal services to the standard of a competent lawyer, and the commentary to that rule was amended in June 2022 to say that maintaining competence means developing an understanding of, and ability to use, technology relevant to the nature and area of the lawyer’s practice — including understanding the benefits and the risks, and recognising the duty to protect confidential information.
The commentary also says the required level of technological competence depends on the practice areas, the geography and the requirements of clients. That is a proportionality test, not a licence to ignore the subject.
The same Rules require a lawyer to hold in strict confidence all information concerning the business and affairs of the client acquired in the professional relationship, subject to narrow exceptions. That duty is broader than privilege — it applies regardless of the source of the information or whether others already know it.
The practical problem is well set out in the Lawyers Indemnity Fund’s notice on generative AI and what lawyers need to know: prompts travel over the internet to the provider’s servers, and a generative system may use what it is given to respond to others. Its recommended questions are the right ones — where is the data stored, who is it shared with, what security applies, will the data be used to train the model, can the material be redacted first, and if not, is client consent necessary?
If a matter is one where privilege will be asserted later, that is a further reason to keep the file out of a public tool. Litigation privilege and solicitor-client privilege are worth losing sleep over in a way that a subscription fee is not.
Ontario’s rules require a lawyer to assume complete professional responsibility for their practice of law, and to directly supervise non-lawyers to whom tasks are assigned. A model is not a non-lawyer, so the supervision machinery is an imperfect fit — but the first half of that obligation is not conditional on who or what produced a draft. Professional responsibility is not delegable, and it certainly is not delegable to a vendor.
Where the firm is on the hook, vicarious liability for a member’s conduct follows the ordinary rules. There is no AI exception.
Your law society is not the only body with a view. The Lawyers Indemnity Fund’s notice on generative AI records that some courts in Canada have started requiring disclosure of the use of generative AI tools in submissions, and its advice is to check with any court, tribunal or authority on its AI policy before relying on the output.
Practice directions differ by court and change without much fanfare, so confirm the current direction of the specific court before you file. This is not a question to answer from memory — and emphatically not one to answer by asking a model.
The Law Society of British Columbia has made AI a strategic priority and publishes guidance on professional responsibility and generative AI covering competence, confidentiality, information security, fraud, plagiarism and copyright, alongside practice advisors available to discuss specific situations. Its Code of Professional Conduct for British Columbia remains the governing text.
Nationally, the Federation of Law Societies of Canada maintains the Model Code of Professional Conduct that the provincial and territorial law societies draw on. It is a source of consistency, not a licensing authority — your own law society’s version governs.
Three rules cover most of it. First, nothing confidential goes into a tool you have not diligenced and whose data terms you cannot describe. Second, every citation, quotation, figure and date is verified against the primary source before it leaves the office. Third, the file records what the tool did and who checked it.
A firm that can produce that record is in a very different position from one that cannot, regardless of whether anything went wrong.
Do we need client consent?
Your terms of retainer and your confidentiality analysis decide it, and the answer changes with the tool. Where confidential material cannot be redacted before use, the Lawyers Indemnity Fund raises consent as the question to ask — which is a fair signal of where the risk sits.
Is legal research the safe use?
Research is the use with the best-known failure mode: fabricated authority. It is safe only if you read every case and provision returned, in the original. Treat AI research as a way to find things, never as a way to know them.
Which law society governs a firm with offices in two provinces?
Both, for the lawyers licensed in each. Where the codes differ, comply with the stricter one; it is rarely worth the argument to do otherwise.
A 30-minute call is enough to tell you whether AI pays for itself here.
